AnyAnomaly enables flexible, customizable video anomaly detection using Large Vision-Language Models (LVLMs) without requiring retraining or specialized expertise.

• Achieves context-aware anomaly detection through natural language prompts
• Eliminates need for environment-specific model development
• Demonstrates strong performance across diverse surveillance scenarios
• Enables immediate deployment in new environments without data collection

This research transforms security monitoring by democratizing advanced anomaly detection technology, making it accessible to organizations without AI expertise or extensive computational resources.

AnyAnomaly: Zero-Shot Customizable Video Anomaly Detection with LVLM